Multi-Segment Polynomial Trajectory Generation of Autonomous Vehicles Based on Quadratic Programming

Trajectory planning is one of the key technologies of autonomous vehicles. The quality of trajectory will affect the driving safety, but the trajectory planning of autonomous vehicles is still a challenging problem. In this paper, a multi-segment polynomial trajectory planning method based on quadratic programming in the Frenet coordinate system is proposed. The trajectory curve is divided into multiple segments to optimize with different emphasis. Firstly, an objective function is designed to consider the moving targets, trajectory comfort, driving safety and acceleration cost. Also, the trajectory continuity, road boundary limit, and vehicle movement limit are considered in the constraints. Next, the optimization problem is solved to determine the coefficients of the multi-segment polynomial, and then the trajectory is generated. Finally, the Gilbert-Johnson-Keerthi (GJK) algorithm is used to detect the collision, which ensures the safety of the trajectory geometrically. Simulation results show that the proposed method can generate a safe and comfortable trajectory. Besides, the proposed method can effectively improve real-time performance and trajectory quality.